结合全局和局部特征的BiGRU-RA图像中文描述模型

针对目前基于全局特征的图像描述模型存在细节语义信息不足的问题,提出结合全局和局部特征的图像中文描述模型.该模型采用编码器-解码器框架,在编码阶段,分别使用残差网络(residual networks,ResNet)和Faster R-CNN提取图像的全局特征和局部特征,提高模型对不同尺度图像特征的利用.采用嵌入了残差连接结构和视觉注意力结构的双向门控循环单元(bi-directional gated recurrent unit,BiGRU)作为解码器(BiGRU with residual connection and attention,BiGRU-RA).模型可以自适应分配图像特征和文本权重,改善图像特征区域和上下文信息的映射关系.此外,加入基于强化学习的策略梯度对模型的损失函数进行改进,直接对评价指标CIDEr进行优化.在AI Challenger全球挑战赛图像中文描述数据集上进行训练和实验,实验结果表明,该模型获得更高的评分,生成的描述语句更准确、更详细.     

A simple texture feature for retrieval of medical images

Multimedia Tools and Applications - Texture characteristic is an important attribute of medical images, and has been applied in many medical image applications. This paper proposes a simple...     

论述建筑装修中的电气施工质量与安全管理

电气工程是科技发展的产物,在先进科技的推动下进步和更新。目前,由于种种原因,电气工程的施工质量和安全存在很多问题。本文主要分析了如何对建筑装修中的电气施工质量与安全进行管理,希望能为相关工作人员提供参考和支持。     

New Rare-Earth Containing （Sr1- yEuy）2 Al2Si10S14O4 Phosphors for Light-Emitting Diodes

Remarkable progress was made in the development of white-light-emitting diodes （LEDs）. White LEDs provided a light element having a semiconductor InGaN light-emitting chip （blue or UV LEDs） and luminescent phosphors. Here we reported the sialon s-phase of （Sr,Eu）2A12Si10N14O4. Eu^2＋ activator ions that were substituted for the strontium site represented a new type of yeUow-green phosphor that could be excited by blue LEDs used for application in the fabrication of white LEDs.     

MHD flow in a rotating vertical cone through a porous medium

In the context of advancements in both heat and mass transfer, the current study intends to analyze the impacts of thermal radiation, Soret, and Dufour on the magnetohydrodynamic boundary layer flow through a vertical spinning cone in porous media. The Dufour effect is the energy flux due to the mass concentration gradient with a reciprocal phenomenon, the Soret effect. Energy expression considers the physical aspects of heat generation and absorption. It is expected that the tangential, circumferential, and normal directions will all have velocity components in flow through a porous media. The governing equations are nonlinear partial differential equations that are rearranged into ordinary differential equations via similarity transformation, and then they are numerically solved using the Runge–Kutta method along with a proper shooting strategy. Graphs are used to examine the impacts of many parameters on flow characteristic velocity, temperature, and concentration, including magnetic parameters, porous parameters, Dufour and Soret parameters, chemical reaction parameters, and more. The numerical findings of the gradient of velocity, the Nusselt and Sherwood numbers, and the surface drag force are tabulated and compared with the current result and the one from the literature. The findings are found to be in good agreement. Circumferential and normal velocities are improved visually for greater values of the porosity parameter, but the tangential velocity behavior of the magnetic parameter exhibits the reverse behavior. In addition, the Dufour parameter and chemical reaction both exhibit diminishing behavior when the Soret parameter increases.     

Bi-and Eu-Doped Y2O3 Phosphors for Ultraviolet Light Emitting Diodes

Red emitting phosphor （BaCa）3 （PO4）2 ： Ce, Mn was prepared by solid state reactions. Effect of chemical component, preparation conditions on the performance of the phosphor was studied by using TGDTG, XRD and phosphor relative brightness meter. It is found that the host （ Bax Ca1-x ） 3 （PO4） 2 is present as a distorted Ca3（PO4）2 structure. Ce^3＋ acts as a sensitiser, and Mn^2＋ is activator. By energy transfer from Ce^3＋ to Mn^2＋, Mn^2＋ is activated and emits red light with wavelength around 650 nm. The phosphor powder shows the highest emitting brightness, when the con- centrations of Ce^3＋and Mn^2＋ are both at 0.15 mol.     

Designing and fabrication of phenothiazine and carbazole based sensitizers for photocatalytic water splitting application

In the present work, we have designed and synthesized two carbazole and phenothiazine donor moieties based metal-free organic sensitizers and their codes are WCBZ2 and WPTZ2 respectively. These sensitizers have been used for photocatalytic hydrogen (H₂) evaluation application. The sensitizers exhibit good light absorption capability and electrochemical properties as well. For increasing water splitting capacity, incorporate platinum salt on TiO₂ semiconductor photoanode was performed and compared hydrogen evolution with pure TiO₂ photoanode. We have also studied the influence of the sensitizer's concentration and the effect of pH of the medium was explored. Using a theoretical measurement optimized both the synthesized dimer dyes structure geometry and the calculated their HOMO-LUMO energy level. Here also reported optimized pH and concentration of sensitizers in the reaction medium and found that the high hydrogen generation efficiency from water splitting is 138.3 μmol (348 TONs) by the WPTZ2 dye.     

基于多尺度残差神经网络的阿尔茨海默病诊断分类

提出多尺度残差神经网络(multi-scale resnet, MSResnet)。采用不同大小的卷积核对图像进行多尺度信息采集,并对神经网络进行残差学习,避免网络退化。对核磁共振图像(magnetic resonance imaging, MRI)进行标准化处理,利用MSResnet模型在阿尔茨海默症(Alzheimer's disease, AD)和正常受试者(normal control, NC)获得的分类准确率为99.41%,在AD和轻度认知障碍(mild cognitive impairment, MCI)获得分类准确率为97.35%。与已有的算法相比,本研究提出的算法的分类准确率得到了明显的提高。     

港产城融合发展理念下的深圳机场临空经济发展规划探索

临空经济已成为城市发展的重要驱动力.随着深圳机场客、货运能级的快速提升,临空经济已进入快速发展期,但港产城不协调问题突出,临空经济发展阶段落后于机场能级.文章基于深圳机场周边高度城市化、高度建成的特点,从港产城融合视角出发,分析了深圳发展临空经济面临的问题,并根据临空经济的时空特性划定临空经济区,明确临空经济区的发展目...     

GZMH：用于有丝分裂细胞核检测和分割的乳腺癌病理图像数据集

目的 有丝分裂细胞核计数是乳腺癌诊断和组织学分级的3个重要评分指标之一,基于深度学习的自动检测方法,可以有效辅助医生进行乳腺病理图像有丝分裂细胞核识别和计数。而当前研究中的公开数据集多为竞赛所用,由举办方联合数据提供者挑选而来,与医院临床应用中所使用的数据存在较大的差异,不利于模型性能及泛化能力的测试验证。针对以上问题,本文发布了来自中国赣州市立医院临床环境的数据集GZMH （Ganzhou municipal hospital）。方法 整理并公开发布的数据集GZMH包含55幅全视野数字切片（whole slide images,WSIs）临床乳腺癌病理图像,提供了用于有丝分裂细胞核目标检测和语义分割研究的两种标注,并由2名高年资医师对3名初级病理医师的标注进行了复核。5种主流目标检测方法和5种经典分割方法在GZMH数据集上进行了训练和测试,检验它们在临床数据集GZMH上的性能。结果 目标检测方法实验结果比较中,SSD （single shot multibox detector）模型取得了最佳的效果,F1分数为0.511;分割方法实验结果比较中,R2U-Net （recurrent rsidual convolutional neural network based on U-Net）性能最佳,F1分数为0.430。所有方法在面对较大规模的临床数据集GZMH时体现的性能都明显低于它们在一些公开数据集上的性能。结论 本文所提出的GZMH数据集能够用于有丝分裂目标检测与语义分割研究任务,且此数据集中的图像更加接近实际的应用场景,在推动乳腺病理图像有丝分裂细胞核分割的研究和临床应用方面具有较大的价值。数据集的在线发布地址为：https：//doi.org/10.57760/sciencedb.08547。